CREDIT AVAILABILITY, BANK CONSUMERLENDING, AND CONSUMER DURABLES

John V. Duca

October 1995

RESEARCH DEPARTMENT

WORKING PAPER

95-14

Federal Reserve Bank of Dallas

This publication was digitized and made available by the Federal Reserve Bank of Dallas' Historical Library (FedHistory@dal.frb.org)

CREDIT AVAII,ABILITY, BANK CONSI]I.IER LENDINC, AND CONSI]IIER DURABLES'

John V. DueaResearch Department

Federal Reserve Bank of DallasP . O . B o x 6 5 5 9 0 5

Dallas , TX 7 5265

Bonnie GarrettDivision of Monetary Affairs

Board of Governors of the Federal Reserve SystexltrJashington, DC 205 51

October L99 5

tWe would l ike to thank, without inplicating, Jean Zhang, Marilda Horvarh, andJeremy Nalewaik for providing excellent research assistance; and Ken Enery,Evan Koenig, David Reifschneider, and two anonynous referees for providingsuggestions on earlier drafts, The views expressed are those of the authors,and do not necessarily reflect those of the Board of Gowernors of the FederalReserwe Systen, the Federal Reserve Bank of Dallas, or the Federal ReserveSysten . Any er ro rs o r o rn iss ions are so le ly ours .

CREDIT AVAII,ABILITY, BANK CONSI'I{ER LENDINC, AND CONSII}TER DUBABI.ES

Abstract

This study tests the ernpirical inplications of a rnodlfied screeningmodel of lending [St ig l i tz and Weiss (1-981, Par t IV) ] us ing a proxy fornonrate credit conditions based on Federal Reserve survey data. Consistentwith screening rnodels, this proxy (1) signlfieantly affects bank consumerlending, (2) is significantly affected by the real federal funds rate and exante default risk measures, and (3) substantially affects consumer durables.Other tesults indicate that deposit rate deregulation has reduced the irnpactof monetary policy on consuner credit availability and consumer durablespend ing .

JEL C lass i f i ca t ion Codes : E5L, GLL, DLz

1. Introductlon

Ernpirical studies of household liquidity or credit constraints generally

fall into two groups. The first tests whether the time series properties of

consumer spending accord with the life-cycle (LCH) or pernanent income (PIH)

h y p o t h e s e s ( e . g . , H a l l ( 1 9 7 8 ) , F l a v i n ( 1 9 8 1 ) , a n d W t l c o x ( 1 9 8 9 ) ) . A s l ^ I i l c o x

(1989) notes, such studies finding violauions of the LCH/PIH cannot determine

whether the violations owe to nyopia, borrowing constraints, or transactions

costs. Partly to avoid this problem, the second stfand of research tests the

LCH/PIH us ing cross-sect ional data on consulpt ion (e.g. , Hal l and Mishkin

(1982) and Zeldes (L989)) or on whether credi t constra ints af fect debt Ie .9. ,

Duca and Rosenuhal (1993) , Jappel l i ( l -990) and Cox and Jappel l i (1992)1.

Hortever, there have been few time series tests of borrowing constraints

on consumer debt . One except ion is F issel and Jappel l i (1990) who est imate

lhe share of credit constrained U.S, households across Llne using parameter

est imates f rorn Jappel l i 's (1990) cross-sect ion study. F issel and Jappel l i

(1990) note that nost of the t i rne ser ies var ia t ion that they est imate ref lects

ehanges in household denographics and balance sheets because they assule that

credit standards prewailing in Jappelli 's (1990) prevailed throughout their

sample. Accordingly, they qualify their results by mentioning r^rhy credit

standards nay vary over time in ways that they may not have ueasured.

This gap in the literature is significant for three reasons. First, it

is important to knorr how much changes in lending policies affect household

borrowing for monitoring the economy and assessing the empirical relevance of

lending policy changes. Second, in evaluating the uransmission channels of

macro policies, it is irnportant to determine the inpact of policy on nonrate

credit conditions. Third, in assessing the relevance of lheoretical models of

credit constraints, it ls important to test rrhether lenders actually react to

changes in the eeonomic environment in nays that are consistent with theory.

This s tudy addresses each of these issues. Section 2 drar,rs out testable

i rnpl icat ions of a uodi f ied St ig l i tz and Weiss (198L, Par t lV) screening nodel

of lending. Section 3 tests irnplications for r,rhat deteanines the availability

of consumer loans frorn banks, while section 4 tests implications for consumer

loan growth at banks. Sect ion 5 assesses the inpact of nonrate credi t

conditions on consumer durables and the conclusion interprets our findings.

2, A Styllzed Theoretieal Mode1 of Bank Consuner Lendlng

This sectlon dewelops a stylized nodel of bank consumer lending similar

to the " redl ln ing" /screening rnodel of St ig l i tz and Weiss (1981, par t IV) .

This rnodel is used to draw out testable implications about \uhat determines che

nonrate terns of bank consumer credit and the volume of bank consuner lending.

Assue that banks are risk neutral and that household j is approved for

a loan of fixed size ($1) if expected loan revenue (Ej=qjR) for some interesr

r a t e ( r ) a t l e a s t c o v e r s t h e b a n k ' s c o s t o f i n s u r e d d e p o s i t s ( d ) , i . e . , :

i f Er=qrR > $ for sorDe R, then household j can qualify for a loan, (L)

ruhere R = (1+r) , q = expected loan qual i ty = ( l -x) , x = expected defaul t ra te,

d = (1+i), and i = riskless real rate. Further assume that the expected

defaul t ra te is inversely re lated to household j 's expected incorne (Byj ) :

x' : ER/ Py1 ,

where g = a positive scaler reflecting the degree of rnoral hazard effects for

a Siven level of (R/Fyj), yj = the household specific component of household

j's expected income, F = a positive index of expected rnacroeconornic conditions

that affect each household's expected income, and yi is distributed uniforurly

( 2 )

over the interval [y*',/ '*] . For tractability, also assrue that the

probability of applying for a loan is constanc across income types.

Under these assr-rmptions, a household j qualifies for a loan if EJ:qiR >

{ for some R, and competition ensures that each qualifying income type faces

different rninimum loan rate. Since loan revenue is quadraEic in R, two

conditions must hold for households having the lowest level of income

prospects (y") that can qualify then fof a loan at some loan rate factor (R"):

q"R" : d '

a ( q " R " ) / 8 R ' : 0 .

Eq. ( 3a) ensures zero

expected loan revenue

The second condition

and ,

expected

curve of

irnplies a

( 3 a )

( 3 b )

prof i ts , whi le eq. (3b) ref lects that the

narginally qualified households peaks at R'.

maxirnurn fatio of debt-payments-to- incone :

P"/ $y') > L/2c, ( 4 )

which is qual i ta t ive ly consis tent wi th typ ica l lending pract ices.

Substitution yields expressions for the rninirnurn level of qualifying

income and the share of households that qualify (6) for a loan, respectively:

y' : 4g$/B, and

6 - ly*"-(4e6/P)l/ ly^* - y '* l , ( 6 )

where 6, < 0, 6u ) 0, and 6" < 0. The first two derivatives reflect that the

"bindlngness " of a given maximuu debt-palrments- to- incone ratio endogenously

depends on real rates and general macroeconomic conditions. 6" reflects the

extent to which a change in other factors not reflected in teaL rnarket

interest rates or in incorne affects credit standards and availabil itv.

+

The degree to which credit avaiLability affects bank consumer lending

also depends on loan demand. Tf the probability of applying for a loan by

household i can be expressed by t(6,a), where fa(0 and fr>0, then the volume

of lending reflects lhe denand of households who are not credit constrained:

L" - o1y*- -y- l r (p ,6) ,

aL/ad - - laet/Fl + fr6 [y*-y'*] < 0,

aL^/aP : beft/F'l + fr6 [y*"-y'i"] ) 0, and

a t^ /as : -g6e t /p l < 0 ,

( 7 )

(8 )

(e)

(10 )

since f r< 0 and f , ) 0 . L"" ( 0 because a r ise ing induces a t ightening of

credit standards. Changes in the macroeconomic outlook (B) and real interest

rates ({) hawe similarly signed supply and denand effects on loan volume. For

example, the second term of (8) is negatiwe reflecting that loan dernand falls

as loan rates rise and the first term is negative since fewer households could

qual i fy for a loan as real ra tes r ise. In equat ion (9) , the f i rs t term is

posiEi.ve because more households could quaLify for a loan and the second term

is positive since loan dernand rises as the macroeconomic outlook improves.

Thus, macroeconornic variables typically included in constuner loan regressions

wi l l a lso p ick up a t leas t some o f the e f fec ts changed c red i t ava i lab i l i t y .

a credit availabil ity proxyIt follows that the uarsinal infornation

o f

in

(CR) for explaining bank consuner loans depends erit ically on how much changes

in CR (1) reflect changes in rnoral hazard problens in lending (Ag), and (2)

respond differently to changes in interest rates or cyclical conditions than

do changes in the demand for durables. In practice, the first category could

reflect changes in bankruplcy laws or the perceived moral hazard posed by

lending to households, and che second could include changes in credit

)

availability stenming froro the bindingness of Reg Q ceilings whose correlation

with interest rates has changed dramatically since deposit deregulation.

In terms of lestable empir ica l hypotheses, eqs. (7) , (9) , and (10) inp ly

that a variable which is positively related to nonrate terms of bank consuner

loans should hawe a posilive correlation with the wolune of bank consumer

lending. In addition, eq. (5) irnplies that such an increasing neasure of bank

credi t avai lab iL i ty should increase ( i .e . , y" fa l ls ) as the real in terest rate

({) falls and as the macroeconomic outlook (B) irnproves. Alternatively, the

standard portrayal of che money nultiplier process also suggests that real M2

growth at banks could be a rnajor detenninant of credit standards at banks.r

3. An Empirieal Model of Nonrate Credlt Conditions at Banks

This section presents a proxy for nonrate credit conditions at banks,

and then discusses its potential deterninants. Next, data used in roodeling

this proxy are presented, and then regression results are reviewed.

3.A. A Proxy for Nonrate Credlt Conditions at Banks

To rneasure nonrate credit conditions, an index of the change ln bank

willingness to nake consumer installment loans is created based on the Federal

Reserve's Quarterly Survey of Bank Lending Terrns, which asked since 1966:Q3:

"How has your bank's willingness to make consumer installment loanschanged relative to 3-nonths ago?"

(a) rnuch more, (b) somewhat rnore, (c) about unchanged,(d) somewhat less, or (e) rnuch less.

A diffusion index (CR) of the average response was constructed by weighing

responses of "much more" by 2, "somewhat nore" by 1, "unchanged" by 0,

"somewhat 1ess" by -1- , and " rouch less" by -2, Posi t ive va lues ind icate

- fo rc ror robanks . Howewer ,

balance theoriesgood daua on ex

imply a role for relative ex ante returnsante returns are unawailable.

o

expanded credit availabil ity and negative values, the converse. In general,

the w l l l ingnes s - to - lend index dec l ines pr io r to recess ions (F igure 1) ,

Because CR tracks xt.e reTative change in wil l ingness to make loans, it

is appropriate to use an equation for the reTative change in bank credit

s tandards to der ive a tes tab le e rnp i r i ca l mode l o f the war iab le CR. In (5 ) ,

the tern y'can be loosely interpreted as a credit standard where a higher

leve l o f y " cor responds xo a tougher c red i t s tandard . Thus , eq . (5 ) imp l ies :

^ los (y " ) : a log (g ) + A los ( { ) - a log (p ) ,

cR : h IA log (g ) , A Iog (d ) , A Iog (B ) ] ,

' : Untll 1979, banks were asked abouttheir credit standards on business loansquestions hawe a significant correlation

and

hor.r their wil l ingness to nake andhad changed. Indexes fron theseo f - . 8 8 f S c h r e f t a n d O w e n ( f 9 9 f ) ] .

( 1 1 )

(L2)

where horo","1 and ho.o"161 ( 0, and h'os(r) > 0. Since CR i-s inversely related to

Alog(y") , (11) impl ies that CR decl lnes wi th an increase in the real r isk less

rate (Alog(C)>0 or a t ightening of nonetary pol icy) , but r ises on s igns of a

better economic outlook (Afog(B)>0, irnplying less nacroecononic default risk).

One problen is the arnb lguity of the survey question used to construct

CR. If banks are nore willing to 1end, (i) hawe they have eased ctedit

s tandards, ( i i ) do more households meet a f ixed credi t s tandard, or ( i i i ) are

banks accommodating an increase in loan dernand? Possibility (ii i) is

ineonsistent with two facts, First, less ambiguous questions about business

lending suggest that the index measures changes in consr-rmer credit standards.2

Second, Granger test resul ts below favor explanat ions ( i ) and ( i f ) .

3 .8. Uodel SDeci fLcat lon and Dara

Severa l regress ions were run. Para l l -e l ing (12) , each inc ludes a measure

of monetary policy and default risk, along with three variables controll-ing

oU'

(dq)(Do

0)U'

(dq)c

<_

'

EoEos(E

.=Itooo

o=

trF

!=2-'

otrc

oo.=oo)cGC)o.FGoE

@(o

sN

<\lt

I

for Reg Q [Jaf fee and Rosen (1978) ] and other regulatory ef fects .

Real Rates and Monetaty Policy Indicators

Four monetary policy indicators were tried. One proxies the change in

real bank funding costs with the change in the real federal funds rate (ARFF),

which outperformed the change in the real l-year T-bl1-1 rate. AnoLher is the

t\ro-quarter growth rate of real M2 balances at banks (GM22).3 Both variables

measure inflation expectations using the annualized two-quarter percentage

change in the implicit consurnption price deflator. Two other indicators of

xhe change in monetary policy that Ire use are t]ne change in the spread betrreen

the six-nonth prime conmercial paper and Treasury bill rates (APAPERSILL)

[Bernanke (1990) and Friedman and Kuttner (L992)] and the change in lhe slope

of the yield curve, defined as the spread between the L0-year and 3-nonth

Treasury rates (AYCURVE) lstock and Watson (1989)]. AFF and PAPERBILL should

be negatively correlated with the change in bank willingness to make consumer

Ioans, while GM2 and AYCURVE should be positively correlated with the index.

DefauTt Risk

Default risk is measured by the percentage change in the quarterly

average level of the index of leading economic indicators (GLI). Because this

index leads the business cycle, GLI should be positively correlated with the

future abil ity of households to service new debts and therefore, with CR.4

3 The tr,/o-quarter gronrth rate outperformed the one-quarter gror,rLh rate.The first difference of GttI2Z also proved inferior to c1422.

n An alternative credit risk measure, the change in the spread between A-and AAA-rated corporate bond yields, was insignificant in other runs, likelybecause it reflects the risk of firms more than that of households. Thechanges in the unenployment and the consumer loau delinquency rate were alsotried. Although these wariables are more specific to households than cLI ,they lat the business cycle and are not be good indicators of ex ante defaultrisk. This likely explains why GLI proved superior to these alternatives.

I

ReguTatory Factors

To control for Reg Q induced dis interroediauion, a variable (REGQ) was

included that equaled the spread of a rnarket rate over a deposit rate ceiling.

Measuring these effects raises the issues of which deposit rate to use and how

to handle market-rate based deposits introduced before the lifting of most

deposit rate ceilings in 1983. REGQ was based on regulations on srnall tirne-

Iike deposits, because their naturity is closer to that of auto loans than

that of demand or passbook savings deposits, and because nost market-rate

based instruments were designed to substitute for small tlrne deposits. REGQ

equals the maximum of zero or the spread between the 3-year Treasury yield and

the cei l in t on 3-year srnal l t i roe rates up through I978:Q2. Frorn 1978:Q3-

81- :Q3, rate ce i l ings on smal l saver cer t i f icates were used fsee Mahoney, et

a L ( 1 9 8 7 ) a n d D u c a ( 1 9 9 5 a , 1 9 9 5 b ) 1 , a n d R E G Q e q u a l e d 0 a f t e r 1 9 8 1 : Q 3 . 5

Another deregulatory action was the introduction of MMDAs in late-L982,

which enabled banks co offer a close substitute for monev market rnutual funds

(MMMFs). As a result, households shifced funds into Uf,fOa" p.ttfy from MMMFS,

5lhis rneasure outperforned others based on either 6-month small tirnedeposi t or noney market cer t i f icate (MMC) bank regulat ions. IQual i ta t iveresul ts wete s imi lar us ing MMC regulat ions. l Ower 1979:Q3-Q4, bank SSCceilings equaled xt.e 2-1,/2 year Treasury yield minus 25 basis points. InJanuary and February 1980, this spread was widened to 75 basis points. FromMarch to April 1980, SSC ceilings equaled the 2-L/2 year Treasury yield minus25 basis points up to 12 percent . Fron June L980 to Ju ly 1981, SSC cei l ingsequaled Ehe 2-L/2 year Treasury yield rninus 50 basis points when this yieldwas between 9.5 and 12.0 percent , 9 .5 percent when th is y ie ld was below 9.5percent , and 12.0 percent when th is y ie ld exceeded 12.0 percent .

The l-ifting of ceilings on uninsured large time deposits in 1973 was ofIinited help in allevlating Reg Q effects for two reasons. First, back in theL970s, it was difficult for srnaller, less well-knom banks and thrifts toissue urrinsured deposits. Second, when rate ceilings r.lere binding on insureddeposits, banks flooded the market with uninsured CDs in periods when monetarytlghtening boosted default risk. As a result, the risk premium that investorsdenanded on large CDs typically soared above the then normal premiurn of 50basis points above Treasury rates. For exarnple, when the funds rate peaked inJrly 1974, six-month CD-T-bill rate spread ,,tas 402 basis points.

9

the supply of funds to banks rose, and banks increased their willingness to

make consumer loans (Senior Loan Officer Opinion Survey, May 1983). A dunrny

(MMDA) equal to l- in 1982:Q4 was included to control for this re interrnediation

effect. Another duflny (coNTRoL) was included for the imposition and lifting

of credi t contro ls in 1-980:Q2 ( : l ) and 1980:Q3(: -1) , respect lve ly , which

caused a tenporary plunge followed by a temporary j urnp in CR.

3,C. Granger Test Resul ts

Granger tests were done to assess whether bankers were confusing loan

dernand r,zi th loan supply in responding to the survey. If banks confused loan

dernand ruith supply, then CR should nove with or lag behind changes in the

growth of real credit extensions and real consurDer durable spending. However,

Granger tests indicate that CR laads both variables (table 1). Granger tests

also show that CR leads changes in the real federal funds rate and the growth

rate of real M2, but suggest that CR moves contemporaneous ly with the paper-

bill spread and changes in the index of leading economic indicators. The

results for M2 may reflect that M2 responds with a lagged response to changes

in its opportunity cost, consistent with evidence that ARFF leads GM22 fand

Alog(M2)1. Tests a lso ind icate that changes in the s lope of the y ie ld curve

Iead CR. 0n balance, cranger tests inply that changes in the willingness-to-

lend index reflect changes in loan policy (supply) more lhan Ioan dernand.

3,D. Results Frour llodeling the lli l l ingnes s-To-Lend Index

Models 1-4 in table 2 include all three regulatory terrns, the ex ante

default risk proxy, and one real rate or monetary policy variable. Models 5

and 6 include ARFF and either AYCURVE or APAPERBILL. Each rnodel has an R'? of

around .80 and several patterns energed. The default risk and regulatory

variables were significant with the expected signs, with a large coefficienu

10

on CoNTROL. The REGQ results irnply that the lifting of Reg Q has reduced, but

not eliminated, the inpact of monetafy policy on nonrate credit terms.

Each nonetary pol"icy wariable lras significant with the expected sign.

Of the models using one monetary variable, uhat using AYCURVE had the highest

R'?, followed in order by ARFF, APAPERBILL, and, GYt22. ARFF is significant in

the presence of APAPERBILL aud AYCURVE. Howewer, uodel 4, whLch uses AYCURVE,

does not track the 1-ending index weII in the early-1990s.

Moreover, unlike ARFF and GM22, AYCURVE and APAPERBILL have statistic-

ally different effects after the 1970s as shown by variables interacting then

with a duIluly equal to l- after 1978 (table 3) whether or not REGQ is present.

This likely reflects a shift in the indicator properties of the spread

wariables.6 By contrast, only when REGQ is onitted does ARFF hawe a

significantly smaller effect after the 1970s, implying that interest rates

affect constrmer credit availability less after deposit deregulation.

In the late 1980s and early 1990s the willingness-lo-lend index turned

down afuer a long period of positive readings (Figure 2), but did not fall as

nuch as it had in previous periods that have been described as credit crunches

(L974-75 and 1979-82) . ' Regression resul ts us ing the real federa l funds rate

imply that uhe absence of Reg Q effects likely accounts for this difference.

6 These findings are consi.stent with evidence of shifts in the relation-ship between the yield curve and inflation [Balke and Enery (1994b)], theyield curve and output lBalke and Ernery (1994a, footnote 7)], and the paper-bilL spread and output [Hafer and Kutan (]-992) and Emery ( forthcorning) I .

The effect of Gl,l22 is not different after the 1970s, reflecting thatirnplicitly by construction, M2 growth picks up Reg Q effects.

? The surveys also showed that many banks tightened their credit stan-dards on C&I and cornmercial real estate loans over L990-91, perhaps oTring toower-borrowing by firms and overbuilding, with surveys showlng that C&I loanstandards were tightened less than during the t974-75 and, !979-82 episodes.

11

4. Uodellng Consumer Loan Growth at Banks

This section tests I,rhether a noninterest rate proxy for credit

availability affects bank consumer lending. First, an eropirical model is

presented. Next, the data are described, and lhen results are presented,

Before proceeding, it is inportant to note that several factors make it

is easier to assess the imDact of credit constraints on bank consumer Loans

than on olher loans. First, one can directly adjust bank consumer loans for

securitizations , which remove loans from bank balance sheets (see Duca and

Carret t (L992)) ; by contrast , i t is nore d i f f icu l t to adjust bank conmerc ia l

and industrial (C&I) or real estate loans for loan sales. Second, tax reform

effects on consunel loans are more transparent and easier to rnodel than those

affecting business or real estate finance. Third, data on auto loan rates at

banks and finance companies provide a good measure of nonbank competition that

is not plagued by tern structure problems posed, for example, by substitution

actoss C&I loans, cornmercial paper, and bonds. Fourth, consumer lending were

not affected by special factors such as those affecting C&I loans (mergers in

the 1980's) or real estate lending ( the thr i f t cr is is) . F inal ly , surveys can

be used to construct a conti"nuous credit availability series tha! goes back !o

the late 1960s for consumer loans. but not for C&I or real estate loans.

4,A, Emplr lca l Spec l f icar ion

This sub-section lays out a sirnple nodel of bank consumer

can control for nonbank competition, nonrate credit terms, and

a1l conslrmer loans. The stock of real bank consurner loans (Lb)

loans which

the deoand for

is g iven by :

Lb : @bQNf , ( 13 )

where oo = bank share of consumer loans, Q = population share qualifying for

12

Ioan large enough that households borrow, lrl = population, and Lt = real

notional consumer loan dernand per capita. Superscripts b and t denote bank

and all lender variables, respectively. A rise ln CR lmplies that more

households can borrow against their permanent income or that Q is higher. CR,

which proxies for the percent change in credit standards, can be seen as a

determinant of how quickly the overall stock of loans adJusts tovrards the

Ievel desired by households based on pernanent income and interest rates.

Under th is in terpretat ion, eq. (13) inp l ies the error correct ion model :

A1" : [c" .+ Er., orCR.-r]ECE-1+ Aabr + Aqr + Ant + A(proxy for 1'"), ( 14)

whete lor,rer case letters denote Logs, A is the fitst difference operator, and

EC =(lb-lb.) is an errot correction terxn reflecling the long-run relationship

betrteen lb and factors affecting loan market share (ob) and dernand (I").3

The Lnteractive variable [CR*EC], tests to see whether relative changes

in credit avallability can affect the speed of adjustnent. Recall that the

error correction tenn is based on the Long-run relationship between the logs

of loans and income, along with the gap betTreen bank and finance company auto

loan rates. Thus, the error correction terD equals the log of the ratio of

the actual co the log of the long-run equilibrium level of loans. As a

result, a tern lnteracting CR and the error-correction terD can be interpreted

as reflecting how short-run changes in credit availability can lead to short-

run changes in how rnuch households are using leverage.

4. B. Data and Var iables

The dependent variable is the growth rate of real bank consumer loans,

adjusted for securitizations (Federal Reserve data) and deflated by the PCE

s Population was excluded because changes in it occur too gradually.

deflator. Independent wariables

tax, spending and labor, nonbank

Real User Cost of Capital

The real user cost of bank

13

are of f ive types: user

conpetition, and nonrate

cos t o f cap i ta l ,

c red i t cond i t ions .

consurner loans (r) can be rneasured by:

( ls )

where R = nomlnal interest raxe, t = marginal income tax rate, n a percent of

consumer loan interest that is deduccible, and ?re = expected inflation. The

four-quarter percent change ln the PCE deflator is used to proxy n".e The

most conmon 48-nonth rate on bank loans for new autos (Fed data) is used

because (a) auto Ioans comprise 40* of consumer borrowing, (b) credit card

debt appears to be very interest insensitive and credit card rates ate \reTy

st icky (Ausubel (1991)) , and (c) much credi t card debt is f loat ( i .e . ,

convenience card use).10 The auto loan rate is adjusted for a break in 1983

when the loan xlaturity changed frorn 3 to 4 years by assuming that each year of

loan maturity adds roughly one percentage point based on anecdotal ewidence.

The tax rate used was the U.S. Treasury's series on the marginal incone tax

rate for a farnily of four earning the median level of incorne (Lerman (1991)).

The share of deductible consumer loan interesc (O) equaled 1 up through

1986:Q4, and since then, the four-year ahead expectation of the average pre-

announced share of interest that is deductible, adjusted for anortization.ll

' This deflator yielded better results than the CPI or core CPL

'o Despite high real loan rates, credi! card debt grew rapidly in the1980s. Other runs using the most corulon credit card rate and the one-year T-bill rate yielded much ruorse Fs and did not affect the qualitative results.

11 The percent of consumer loan interest that could be deducted r,ras 65*for L987, 40* for 1988, 20t for l -989, lOt for 1990, and 0 thereaf ter . Thefour-year horizon xnatches the rnaturity of the bank auto loan rate series.

r - R ( l - t O ) - ? r e ,

L 4

VariabTes ControTTing for Other Tax Effects on Bank Consuner Loans

Aside frorn changing lhe after-tax real rate via altering interest

deductibility, the Tax Reform Act of 1986 also changed refative cost of

consulDer loans versus other finance, inducing many to pay off consurner debt

with inueres t-deductible horne equity lines of credit (HELCs). To control for

this substitution effect, several dunmy variables were tested including a

dummy equal to I since f986 (REFDIJM), a dummy egual to L in the transitional

quar ters of 1986:4 and 1987:1, and a dumny equal to I in the t ransi t ional

quarter of L981 :L. All of these tax reform duuunies were insignificant and did

not affect the qualitatiwe results in the error-correction models tested, and

are not in any of the models that are reported in che tables or charts.lz

Consumer Spending and Unenp Toyment

Several other variables were included that are typically associated with

Ioan demand, Because loans reflect a deriwed demand for durable purchases,

the Federal Reserve Board model's proxy for permanent incone (CON) was

included to control for the underlying demand for durables. Other demand

effects were controlled for with the change in the unemploynent rate (AU).13

ControTTing for Substitution Between Bank and NonBank Consumer Loans

Auto loan rates at banks greatly exceeded those at finance companies

until the late-1970s. Following the introduction of market-rate based deposit

instruments in the early-1980s, this spread narrowed as banks passed on the

'2 Sone of these durnmies were significant in other nodels that dld notinclude long-run relationships. Most households with HELCS init ially usedthen !o re t i re eonsumer debt [see Canner , Lucket t , and Durk in (1989) , p . 337] .

IrCON = consulption of services and nondurables plus the imputed serviceflow frorn the stock of durables. CON awoids much of the simultaneity betweendurable speuding and loans. Loan supply effects of CON are l ikely picked upby CR, implying that any marginal information in CON reflects loan dernand.

l 5

h igher deposi t ra tes and banks lost auto loan rnarket share [Duca (1991) ] .

Since then, the captive finance companies hawe irregularly offered incentives

which induced shifts toward auto finance coupany loans [Duca (1991)]. To

parsinoniously control for these effects, we use the spread between finance

and bank auto loan rates for nerr cars (SPREAD). A bigger spread should raise

bank loan rnarket share (ao) and bank consurner loan growth, (Af').*

Non-Rate Cred.it Condit ions / Avai Lab i7 itv VariabTes

Two variables were used to control for the effects of variation in non-

rate credit conditions/availab il ity. The first is CR. Thls index, however,

is more tlghtly linked to the gror,/th rate of real consumer Ioan extensions

than to the growth of consumer loan outstandings, owing to aoortization,

principal payments are back-loaded over a loan's life, and loan outstanding

growth tends to lag loan extension growth by about one year. Unfortunately,

extensions data end in 1982:Q4, necessi ta t lng the use of outstandings.

We tested the one-quar ter lag of CR as a separate r .h .s . war iable and in

lhe form of an interactive variable that equaled the one-quarter lag of CR

multiplied by the error-correctlon tern. The CR wariables control for two

effects. First, they reflect credit standards at banks and perhaps at

nonbanks. For this reason, an increase in CR rsil1 likely accornpany faster

loan gtowth [higher Alb in eq. (2)] beeause the share of households that can

qual i fy for a loan r ises (Aq is h igher) . Second, a r ise in CR may lead to

banks gaining market share from finance companies [see Duca (1991)]. One

reason for thls is that banks had tlghtened credit standards when binding Reg

1a The finance conpany rate is an average on all loans, while the bankra te i s fo r a f i xed na tur i t y (3 years be fore 1983 and 4 years a f te rward) . (Feddata). SPREAD was adjusted for maturity differences by assuming that eachextra year of loan maturity boosts loan rates by one percentage point.

L 6

Q ceilings caused deposiE outflows while finance companies continued raising

funds by issuing conmercial paper.

Another reason a rise in CR boosts bank rnarket share is related to the

lendet of last resort role of captive finance cornpanies, r,rhere Loan delin-

quency rates tend to be higher. Captives lend to some who are denied credit

at banks because part of the profit earned on a car sale exceeds the negative

expected return from lending to a poor credit risk. When banks tighten their

credit standards, they iDprove the average loan quality of people denied bank

Loans, thereby inducing finance companies to tighten their credit standards

less than banks do. As a result, a fall in CR rnay also be correlated with

falls in Aol and Alb as banks lose market share to finance conpanies.

The second credit availability variable accounts for the consuner credit

contro ls of 1980:Q2 and thei r l inger ing ef fect in 1"980:Q3. These contro ls

induced banks to tighten their credit standards on credit cards. Although the

controls exempted auto credit, they lowered banks' auto loan market share for

two reasons Isee Duca (1991) ] . F i rs t , many consulers bel ieved that i t was

iIIegaI to borrow and may not have borrowed untiL they met dealers r^rho could

originate loans. Second, many consumers may not hawe applied for bank loans

fearing that they would be rejected. To distinguish this effect from other

factors affecting credit availability, a duu[oy variable (DftM802) was included

that equals 1 in 1980:Q2 and that is expected to hawe a negative sign.

4. C, Resul ts

A two-step error-correction model of consumer loans was developed. In

the first step, a cointegrating vector \sas estlmated (with a trend) using the

Johansen and Juselius procedure to find a long-run relatlonship between the

log of real consumer loans, the log of perrnanent income (CON), and the gap

L7

between finance company and bank loan rates (SPREAD).15 The resulting vector

shows that lb is positiwely related to log(CON) and SPREAD over the long-run.

From this cointegrating vector, an error correction tenn (EC) was

defined and used to estixlate second stage model versions of equation (14):

A lb" - [ c " + e rCR._1]EC"_1 + P\zc_ j + l l ^ lb r_1 + . \ r^ lb r - r , (L6)

where z = a vec tor o f shor t - run war iab les , inc lud ing A log(CON) ,_ , , A(SPREAD) . ,1 ,

rr-1, DuM802c, and A(U)r-1, where U: the unemplo)-rrent rate. [Other lags of

these variables were insignificant. ] Except for models L, 3 and 5, which

exclude DUM802, the nodels in table 4 include the same z and EC terms, but

d i f fe r in whether CR is added as separa te r .h .s . te rm or i s ln te rac ted w i th

the EC terrn to see if credit availabil ity is positively correlated with the

speed o f e r ro r -cor rec t ion . S ince loan da ta s ta r t in 1973:Ql and 2 lags o f A lb

are inc luded as regressors , the sarnp le was L973:3-94 :4 .

ECr,1 and A(U)t-l are significant in every model with the anticipated

signs, while the rate term r r^ras ins ignificant .16 Note that while the short-

run te rns A(SPREAD)E_L and A( log(CoN)) " , a re ins ign i f i can t , the s ign i f i cance o f

15 The real after-tax loan rate is stationary and is thus not included,0n1y when SPREAD r,ras included could a significant cointegrating vector befound, reflecting Lhe decline in bank's auto loan market share in the late-1970s and early-1980s when bank loan rates rose toward finance conpany rates.

16 EC should have a negatiwe sign as EC= lb - lb's equil ibrium level. Inother runs, r was replaced with log(r) and Ar. These allernatives ruere alsoinsignificant and did not affect the qualiuative results. r was significantin one-stage error correction nodels that, in place of the EC tern in the ttwo-s tage mode ls , inc luded the one-quar te r lags o f log(L) , 1og(CON) , SPREAD, andr . In some one-s tage node ls , CR was added as a separa te r .h ,s . var iab le o rwas interacLed with Iog(L). Slnce the qualitatiwe results with respect to CRwere similar co those in the tables, the more elegant two-stage results arepresented. In other runs, the percent change in consumer confidence (CoNFID)was insignificant, irnplying that confidence dld not add inforrnation in thepresence of other dernand-related variables.

1 8

Iog(CON).-, and SPREAD.-, in the cointe gx^xing vector inplies that incone and

the degree of nonbank coupetition are significant long-run determinants of

bank lending. DIJM802 is significant when included, and its absence in models

3 and 5 has little effect on the significance of CR. The terms interacting EC

with lags of CR in models 3 and 4 are jointly significant. Because the EC

term is expected to have a negative sign, the negative signs of CR*EC in those

rnodels irnply that greater credit availability speeds up the adjustnent of bank

loans to their long-run equilibriurn and encourages people to lever up in the

short-run. Furthermore, in rnodels that include the significant credit control

duumy (nodels 2,4, and 6, ) , thete is autocorre lat ion in the res iduals of the

noncR model (2) but not in nodels 4 and 6 that include the interactive CR term

and CR, respectively. 0n balance, these findings iurply that the growth rate

of bank consuner loans reflects both nonprice and price terns.17

Using an insa-uple per iod of L973:3-87:4, ex post forecasts f rorn the

table 4 models r,rere done over 1988:Q1-94:Q4 (the most recent interest rate

cyc le) us ing an insample per iod of L973:Q2-87:Q4. Compar ing across models

with or without the credit control durnrny, the S.S.E. of the forecasts frorn the

nonoR models (models I and 2, respectively) were 28 percent higher than those

from the uodels with the interactive CR term (rnodels 3 and 4, respectively)

and 29 percent higher than those from the nodels with CR (models 5 and 6).

5. Consuner Spendlng Effects

To assess the impact of nonrate credit terns on consrrDer spending, the

credit awailability index (CR) is added to several equations in rhe Federal

17 Findings are consistent with resultsDuca (1991) finds that auto lending shifredcompanies when banks became less wil l ing tofinds cross-section evidence that l iquidity

shown later and two studies,fron banks toward financemake consumer loans. Lan (1991)constra ints af fect auto sa les.

! 9

Reserwe Board 's (FRB) econometr ic model .

auto and nonauto durable spending and the

services and nondurables plus the service

Effects on Consu[rer Durables

Two are error-correction nodels of

third rnodels consumer spending on

flow from the stock of durables.

and

The FRB durable equations are of the form:

i . / k , - , - a E C . - , + B A x . , ( 17 )

where i = durable purchases and leases ninus depreciation, EC.-, = the error-

correction term, and X = a vector of variables having short-run dJmarnic

effects. The dependent varlable irl the auto rnodel is household purchases and

leases of autos divided by the lagged stock minus a constant (.17493) and a

tine trend (l0.398rztine). The dependent wariable in the nonautos model equals

purchases of nonautos diwided by the lagged stoek (there ls no tlme trend).

In the EC terms beloTr, the first tern controls for permanent income (con), the

second for the user cost of capital, and the chird for relative prices:

EC"

EC"

( 1 8 )

(1e )

where "a" ( "o") superscr ip ts denote auto (nonauto) war iables, CON = consumer

purchases of servlces and nondurables plus the imputed servlce flolr froro the

stock of durables, r " = real a f ter -Lax in terest rate for autos adjusted for

relative prices, ro = real after-tax interest rate for nonauto durables, Ps." =

price of gasoline, P"on = consuuption price deflator, MPG = fuel efficiency of

the existing auto stock, and Po = price deflator for nonauto durables.

For autos and nonautos, x includes Arr and the one-quafter lags of

disposable income growth (Ayd) and the dependent variable. For autos, x also

20

includes the percent change in the real price of gas per nile driven

[Ps8/Poi*MPC], and the two-quarter lag of the dependent variable.

Of the models ineluding CR, the nost successful in terms of f it and

perfornancele of the EC specification were of the forn:

i , /k"_, : [oo + o lcR]Ecr_ l + p^xt , ( 20 )

where ao is a constan! and a1 is hypothesized to be positive because increases

in credit availabil ity (CR>O) theoretically could increase the responsiveness

of purchases to the tap between desired and actual durable stoeks. Since the

speed of error correction is more a function of the Ievel rather than the

percent change in credit availabil ity (CR), other runs also included lags of

CR interacted with EC.-,. Giwen the big spikes in CR around the credit control

episode, some models also include the credit control dunny, CONTROL,

intetacted with the EC term to see lf the credit control episode can bias the

nornal dynamic effect of CR on the speed of error-correction.

CR has significant, positive effects on the speed of adjustnent whether

contemporaneous or lagged CR is intefacted vrlth the EC tern (table 5). For

both durables, the one-quarter lag of CR has the largest effect, and x0ore so

when the significant credit control durnrny is present. Consistent with these

results, both FRB models fit better when CR is included as a determinant of

the speed of correction. In addition, including CR in the nonauto durable

model allor^rs one to not reject the null hypothesis that the residuals are

well-behaved. Based on fit and t-statistics, models 3 and 6 are the preferred

speciflcations for auto and nonauto durables, respectively.

'8 In ocher runs, separateest imated EC coef f ic ients rqere

Iags of CR were very signifi-cant, but theinsignificant and R"s were lorrer.

2 !

Ex post forecasts of the levels of auto and nonauto durables frorn the

fRB models hawe S.S.E. 's r rh ich are 408* and 144* h igher than those der ived

from models 3 and 6, respect ive ly , over the post-Reg Q era (1983:4-94:4) . le

Unlike the FRB auto rnodel, model 3 did not incorrectly predict a sharp decline

in auto purchases over the 1988-90 per iod of r is ing in terest rates, d id not

incorrectly forecast a sharp recovery during Che subsequent period of

decl in ing in terest rates over 1990-93 (F igure 2) , 'zo and does a respectable job

of sinulating auto spending out-of-sample over a 12-year period. One

plausible explanation is that the FRB rnodel is affected by bias from omitting

the credit awailability index that lead it to overestimate the effect of

higher interest rates on autos in the post-Reg Q era. Tn particular, credit

awailability, as measured by CR, declined much less during 1988-90 run-up in

interest rates than in preceding periods of Fed tightening. Additionally, CR

did not surge as nuch as in prior periods of Fed easing during the 1990-92

fall-off in interest rates beeause prior increases in the federal funds rate

in the late-1980s did not lead to Reg-Q induced dis intermediation thar was

Iater unwound by the subsequent rate cuts in the early-1990s.

credit Avaitability Effects on trCON"

The variable "CON" plays a critical role in the FRB model of the U.S.

economy because it proxies fof peruanent income and measures consumption in a

way consistent with the LCHIPIH. The lRB model of "CON" is:

le While the CR-augnented uodel of nonauto durables also outperforms theFRB nodel in ex post forecasts, the difference is not as striklng as in thecase of autos. This rnay owe to improvements in computers that make itdiff icult to track nonauto durable dernand well ou!-of-sample and to the factthat because nonauto durable itens are less expensiwe than new cars, nonautodurable purchases are relatiwely less dependent on obtaining financing.

?0 Figure 2 plots forecasted levels of real auto durable purchases thauwere derived from durable stock growth forecasts generated by the models.

rl@

!o LJ

U)

: U'

o c E !P 'Tl

o o o g) o o ct = o o o 6'

o ! tr .l o qt o o th

I

a_

\

T GI c o 19

I

:-)_

: -.

-n ng

wc

o o

_t't o o

:\

22

CON. - l i=e, ra._iyr._i* 5i-0,rp._iyt't_1+ ti=0.36r_1yp"_r* ?w". + OT,e"r+ roll-r, (20 )

where CON, the w's and the y's are per capitar yr=after-tax labor income

(instrumented using Lags of yt and the real federal funds rate), y" = transfer

incone, yp=after-tax property income, I,rs (w")-the six-quarter moving average of

stock market (other) wealth deflated by P"", and OIL = L974 oLL shock dumrny.

Since CR has a rnean near zero, CR nay temporarily affect savings. As a

test , a set of var iab les in teract ing CR, 1 wi th y I . - i were added to eq. (20) .

These terms are jointly significant with a positive sum of coefficients and

residuals are better behaved in their presence regardless of whether jointly

insignificant terms interacting yl.,i with a credit control duumy are present

(cable 6). These results suggest that loosening nonrate credit terns boosts

consumet spending by temporarily lowering the savings rate.

Nevertheless, the interactive terms have a smaller effect on nodel fit

and are less significant in the "CON" models than in the the two durable

models. These d i f ferences p lausib ly ref lect that "CON" inc ludes: (1) non-

durable purchases which are much less credit intenslve than durable purchases,

and (2) the imputed service floru from che stock of durables which is more

stable and less affected by credit conditions than are ourchases of durables.

5. Conc lus Lon

This study finds that bank consumer lending reflects not only demand

factors and supply influences related to interest rates, taxes, and nonbank

conpecition, but also nonrate credit conditions. In addition, a proxy for

nonrate credit conditions is found to have significant effects on consumer

durable spending, These results are conslstent wlth eross-section ewidence

that borrowing constraints have large effects on households [e.g., Duca and

Rosenthal (1993) , Jappel f i (1990) , and Cox and Jappel l i (L992) l and wirh

z3

theoret ica l nodels that bank credi t is a l loeated qr i th nonrate terms (e.g. ,

Jaf fee and Russel l (1976) , St ig l i tz and lJe iss (1981) , and l , I i l l iarnson (1986)) ,

as well as rates. Ewidence also sholrs that bank willingness to make consumer

Ioans is decreasing in the real federal funds rate and positively related to a

leading indicator of better uacroeconomic conditions. These two findings are

consis tent wi th a nodi f ied St ig l i tz and lJe iss (1981, par t IV) sereening model ,

a view that credit availabllity is partly endogenous, and with the continued

use of credit standards by banks in evaluating loan applications. Overall,

the results inply that, in addition to working through interest rate and

income channels, nonetary policy influences bank consuner lending and consumer

durable purchases by affecting nonrate credit conditions. Findings also have

two other inportant implications for monetary policy. First, the impact of

federal funds rate changes on the willingness to lend index is smaller after

the 1970s in rnodels onitting a Reg Q wariable and uodels with such a variable

irnply that deposit deregulation has expanded credit availability. Second, the

onisslon of the lending index likely creates an upr{ard bias to the post-1970s

inpact of nonetary policy, consistent with the fact that the FRB model of

consumer durables severely underpredicts consumer durables in the late-1980s,

while a model including the credit availability index as a determinant of

stock adjustment does not (Figure *). Together, these findings inply that

deposit deregulation has reduced the impact of interest rates on consumer

credit availability and are consistent with the position of Miron, Romer, and

Weil (1995) that the relative strength of the bank credit channel of monetary

pol icy has decl ined ower t ime.

24

References

Ausubel , L .M. , 1991, The fa i lure of compet i t ion in the credi t card narket ,

American Econornic Review 81". 50-81.

Balke, N. and K.E. Ernery, 1994a, The federal funds rate as an indicator of

nonetary policy: evidence fron the 1980's, Federal Reserve Bank of Dallas

Economic Rewiew, F i rs t Quarter , 1-15.

Balke, N. and K.E. Ernety, 1994b, Understanding the price p.uzz1.e, Federal

Reserve Bank of Dallas Economic Review, Fourth Quarter, L5-26

Bernanke, 8.S. , 1990, 0n the Predic t ive Power of fn terest Rates and Interest

Rate Spreads, New England Economic Review Now./Dec. , 51-68.

Canner, G.8. , C.A. Lucket t , and T.A. Durk in, 1989, Hone equi ty lending,

Federal Reserve Bulletin, 333-44 -

Cox, D.C. and T. JappelLL, L992, The ef fect o f borrowing constra ints on

consuner l iab i l i t ies, Journal o f Money, Credi t , and Banking 25, L97-2L3.

Duca, J .V. , 1991, Credi t avai lab i l i ty , deposi t account innovat ion, and thei r

effects on the U.S. auto loan market, unpublished mimeo, Federal

Reserve Bank of Dallas

Duca, J .V. , 1995a, Est inat ing the in terest sensi t iv i ty of U.S. GDP using

accurate Reg Q measures, Federal Reserve Bank of Dallas Research Paper No.

9 5 1 3 .

Duca, J .V. , 1995b, Reg Q and housing coef f ic ients, Federa l Reserve Bank of

Dal las Research Paper No. 9512.

Duca, J .V. and B. GarretX, L992, The ef fects of credic avai lab i l i ty , nonbank

competition, and tax reform on bank consurner lending, Federal Reserwe Bank

of Dal las Research Paper No. 9207.

Duca, J .V. and D.L. Rei fschneider , L992, Non-rate credi t condiu ions and

z5

consumer durable purchases, unpublished manuscript, Board of Governors of

the Federal Reserve Systen.

Duca, J .V. and S.S. Rosenthal , 1993, Borrowing Constra ints , Household Debt ,

and Racial Discrimination in Loan Markets. Journal of Financial

In temediat ion 3, 77-103.

Enery, K.E. , for thcoming, The in format ion content of the paper-b i l1 spread,

Journal of Econornics and Business.

Fissel , G.S. and T. Jappel l i , 1990, Do l iqu id i ty constra ints vary over t ine?,

Journal of Money, Credit, and Banking 22, 253-62.

Flavin, M., 1981-, The adjustment of consumption to changing expectations about

future incone, Journal o f Pol i t ica l Econouy 89, 974-LO09.

Fr iedman, B. l '1 . and K.N. Kut tner , 1992, Money, income, pr ices, and in terest

rates, r t Amer ican Economic Review 82, June 1992, 472-92.

Hafer , R. l , I . , and Kutan, A.M. (1992) , "On the money- lncome resul ts of Fr iedrnan

and Kuttner,'r Southern Ill inois University at Edwardsville Working Paper no.

92-0303 (Edwardsvi l le , IL) .

Hal l , R. , 1978, Stochast ic impl icat ions of the l i fe cyc le pernanent income

hypouhesis : theory and ev idence, Journal o f Pol i t ica l Econony 86, 97L-87.

Hall, R. and F. I'tishkin, L982, The sensitivity of consurnptlon to transitory

income: est imates f rou panel data on households, Econometr ica 50, 461-81.

Jaf fee, D.M. and T. Russel l , 1976, f toper fect in format ion, uncer ta inty , and

credi t ra t ion ing, Quarter ly Journal o f Economics 90, 65L-666.

Jaf fee, D.M. and K. Rosen, L979, l loxtgage credi t avai lab i l i ty and res ident ia l

construct ion, Brookings Papers on Economic Act iv i ty , 333-76.

Jappel l i , T. , 1990, Who is credi t constra ined in the U.S.?, Quarter ly Journal

of Economics L05, 2L9-34.

2 5

Larn, P. , 1991, Permanent incone, l iqu id i ty , and adjustments of automobi le

stocks: Evidence frorn panel data, Quarterly Journal of Econonics 106, 203-

3 0 .

Lermau, A.H. , 1991, Average and rnarg inal income tax, soc ia l secur i ty , and

rnedicare tax rates for four-person fanilies at the sane relative

positions in the income distribution, 1955-91 , unpublished nanuscript,

Office of Tax Analysis, Departrnent of Treasury.

M a h o n e y , P . I . , A . P h r h i t e , P . F . 0 ' B r i e n , a n d M . M . M c l a u g h l i n , 1 9 8 7 , R e s p o n s e s

to Deregulat ion: Reta i l Deposi t Pr ic ing f rorn L983 through 1985, Staf f Study

Number 151, Board of Governors of the Federal Reserve System.

Miron, Jef f rey A. , Chr is t ina D. Romer, and Dawid N. Wei l , "His tor ica l

Perspectives on the Monetary Transnission Mechanism, " NBER Reprint No. 1931,

Januaty 1995.

St ig l i tz , J .E. , and A. Weiss, 1981 , Credi t ra t ion ing in rnarkets wi th

imperfect information, Anerican Econornic Review 71, 393-41-0.

SLoek, J. and M. I,Iatson, 1989, New Indexes of Coincident and Leading Econouic

Indicators, in NBER Macroeconomics Annual 1989, O.J. B lanchard and S.

Fischer , eds. , Carnbr idge, MA: The MIT Press, 351-94.

l^ I i lcox, D.E. , 1989, Socia l seeur i ty benef i ts , consunpt ion expendi tures, and

the l i fe cyc le hypothesis , Journal o f Pol i t ica l Economy 97, 288-304.

Wi l l ianson, S.D. , 1986, Cost ly noni tor ing, f inancia l in termediat ion, and

equilibrir.rn credit rationing, Journal of Monetary Economics L8, L59-79.

Zeldes, S.P. , L989, Consumpt lon and l iqu id i ty constra ints : an empir ica l

invest igat ion, Journal o f Pol i t ica l Economy 97, 305-46.

27

Table 1: Granger Causality Results

Direct ion

CR _> DUR

cR -> AlD

CR -) Aleb

CR -> ARFF

cR -> cM22

F S tat is t ic

L O . 2 2 "

5 . 2 4 ' .

o . u J

4 .16*

4 .71 "

2 . 27

2 .74 '

2 . 00

Direct ion

DUR _> CR

Al' -> cR

Aleb -) cR

ARFF -> CR

cM22 -> CR

F Sta t is t i c

2 .L0

2 .39

1 .33

1 .91

L . 57

CR -> APAPERBILL 1. 98 APAPERBILL _> CR L.7 6

CR -> AYCURVE

RFF -> GM22

CR .> GLT

AYCURVE -> CR

GM22 -> RFF

GLI _> CR

5 . 2 9 . '

. 6 4

1 . 3 0

A1l tests include a durnny for the inposition and lift ing of credit eontrols in1980 (CONTROL) and are based on 4 lags w i th samples over 1968:Q1-94:Q4, exceptfo r d leD where da ta a re awa i lab le fo r 1968:Q1-82:Q4 sample per iod .

.-> denotes test of causality fron the left-hand side wariable to the ridhr-hand side variable.

' ( " ) denotes s ign i f i can t a t the 5 t (1* ) Ieve l .

Var iab le Def in i t ions :CR = index of relative change in bank wil l ingness to rnake

consuner installnent loans.

AIb

AIe"

ARFF

GTI22

= percent change in real bank consumer loan outstandings.

= percent change in real bank consumer installrnent loan extensions.

= first difference of the federal funds rate ninus the annualized 2-quarter growth rate of PCE deflator.

= 2 quarter gror,Tth rate of real M2.

APAPERBILL = first difference of the spread between the 4-6 month primecommercial paper rate and the 5 rnonth Treasury bill rate.

AYCURVE = first difference of the spread between the l0 year Treasurynote and 3 month Treasury bill rate.

= L quarter percent change in the index of leading economicind ica tors .

GLI

28

Resulrs of Modeling an#"":t;"tr.nk wirlingness to Lend(Samp le : 1967 :Q l -1994 :Q4)

con-stant

1

27 .49*(7 .48 )

ARFFt -2.42"( - 3 . 2 8 )

ct422E

APAPERBILLE

AYCURVET

c l r 2 r 3 . 1 4 " ' ,( 4 . 8 2 )

REGQT -10. 75"',( - 5 . 4 2 )

CoNTROI+ -52.36"( - r_0 .64)

MMD,\ 25 .99"(3 . r0 )

2

1 9 . 6 0 "( 5 . 7 0 )

Models

3

2 L . 9 4 " "( 6 . 2 L )

4

20 .86 "(5 .83 )

5

26 . L4"( t .2e)

- 2 .05 "( -2 .1e )

- ) . z )( -2 .4 r )

6

24 .52"( 6 . 6 6 )

- 1 . 6 8 ' ,(-2 . z0)

L . I ' +

( 2 .00 )

- o . . + r( -2 .e5)

2 .2s*(2 .eL)

- 1 1 . 1 9 "( _ 5 . 6 1 )

- s3 .82 ' .( - 10 .46 )

3L .96 . ' ,( 3 .51 )

3 . 40-.( 5 .08 )

-L1" .43"( -s .7s)

-49 .92"( -e .71)

2 4 . 8 9 ' "(2 . e5)

4 . 7 1 . .( 3 . 7 5 )

( 5 . 7 1 )

- l-0 . s3*( -s .28)

-55 .83*( - L1 .80 )

38 .87 "(4 . s3 )

3 . 4 8 - -( s . 38 )

-L0 .27 "( -5 .2e )

-48 .7T . .( - e .63 )

23 . 26 *(2 .81 )

3 .78 - -(2 .80 )

3 .90 "(5 .72 )

-9 .78 - .( - 5 .03 )

-53 .91 "( -1 r . 20 )

34 .83 . .( 3 .98 )

R , . 810 . 797 . 805rho .690 .653 .7O7D. rd . 1 .8s 1 , . 90 L .93Q(24 ) 23 .87 L9 .45 23 .20' ( " ) - -s ign i f i can t

a t the 5 t (1 - * ) 1eve1.

. 815 . 820 . 823

.116 . 684 . 6861 . 90 t . 93 l - . 88

20 .44 2 t . L5 22 .83

t - ra t ios a re in paren theses .

var iable Def in i t ions:ARFF = first difference of the real federal funds rate.CIl22 = 2 quarter growth rate of real M2.APAPERBILL = A of 4-6 mon. commercial paper rate minus 6 rnonth T-bill rate.AYCURVE = A of the 10-vr. Treasurv note-3 month T-bill soread.

= 2 qtr. pereent change in the index of leading economicindicators.

REGQ = measure of b indingness of regulat ion Q cei l ings.CONTROL = credi t contro l dunrny : 1 1n 1980:2 and -1 in 1980:3.MMDA = duxnny equal to L in 1-982:Q4 when MMDAs were introduced.

2 9

Table 3: Impact of Excluding Reg Q Measures onEstimated Coefficients on Monetary Policy Measures

( fu l l sarnple, L967 :L-94:4)

s Abs. Change inMonetary Models Models Monetary Coeff.Variable ruith REGO without REGO Fron Droppine REGQ

A R F F E - 2 . 4 2 " - 3 . 7 8 " - 3 6 *( - 3 . 2 8 ) ( 4 . 4 8 )

cI422E

APAPERBILT€

aYcuRvEr

I . I +

( 2 . 0 0 )

- 6 . 4 L - -

" - t q 5 \

L . O I

( 2 . 6 6 )

- 8 . 1 8 . '( - 3 . 3 7 )

6 . 3 8 "(4 .67 t

-32?

-252

lnteractive Dununy Tests for Different Post-L978 Effect of Monetary Variablesl

MonetaryVariable

ARFFT

cNI22r

APAPERBILI,T

AYCURVET

r .92(1 .30)

-0 .62( -0 .6e)

-r3 .26"( -3 .04)

- 6 . 6 2 '( - 2 . 4 2 )

( 2 . 3 L )

- 0 . 8 4( - 0 . 8 0 )

- 9 . 3 6 .( - 1 . 7 1 )

- 8 . 9 2 "( -3 .0s )

Models Modelswith REGO vrithout REGo

1. T'trese tests add an interactive dunmy variable to each nodel using onemonetary poli.cy measure, where the i.nteractive variable equals the product ofthat monetary policy variable and a durnmy equal to one after 1978:Q4. Themodels correspond to rnodels 1-4 fron table 2, with and rnrichouu REGQ, as noted.' ( " ) - -s ign i f icant at the 5t (18) level . t - ra t ios are in parentheses.

Var iable Def in i t ions:ARFF - first difference of the real federal funds rate.AM22 = 2 quarter gror{th rate of real M2.APAPERBILL = A of 4-6 non. corumercial paper rate minus 6 roonth T-bill rate.AYCURVE = A of the lo-vr. Treasurv note-3 month T-bill srread.

Table 4:

Var iables

constant

F '

cRE_!*ECE_1

CR"-t

DUr'1802.

A log (Lb ) t - r

A l o g ( L b ) " .

Alog(coN)"_1

ASPREADT-1

rt-r

AU.-t

Results of

Model 1

- . 38 5 8 . -( - 4 .30 )

- - u b . I f

( - 4 . L 9 '

Modeling

Model 2

- . J f , 4 6

( -4 .54 )

0 .0s94 ' -( -4 .42 )

0 .5518 . ' ,( 4 .38 )

0 . 1968-(1 .78 )

0 .3503(1 .33 )

- . 0007( -0 .66 )

0 . 0004(0 . s5 )

- . 007 2 '( -2 .52 )

- . 0337 ' -( - 5 .07 )

o .5244--(4 . 6e )

0 .2180 '( 2 .26 )

0 . 303 I( r - . 33 )

- . 0008( -0 .86 )

0 . 003 5(0 .68 )

- . oo7 4 "( -2 .ee)

3 0

Real Bank Consumer Loan

Model 3 Model 4

- _ J J b 6 - . ) Z Z )

( - 4 .08 ) ( - 4 . 3s )

- . 0568 " - . 0s43( -4 .01 ) ( -4 .28 )

- . 0030 " - . 0023( -4 .06 ) ( - 3 .34 )

- . 0285*(-4 .44)

0 .4402" 0 .4378"(3 .6s ) ( 4 . 0s )

0 .2529- 0 .2s74"(2 .48 ) ( 2 .82 )

0. t -433 0.L531(0 .58 ) ( 0 .70 )

- . 0004 -0 .0005( -0 .43 ) ( - 0 .6s )

- . 0002 - . 0001( -0 .36 ) ( - 0 .10 )

- . 0011 ' - . 0068 - -( - 2 .44 ) ( - 2 .8e )

Growth 1973:Q3-94:Q4

Model 5 Model 6

- . 3 4 0 3 " " - . 3 2 5 1 . .( - 4 . 1 3 ) ( - 4 . 4 0 )

- . 0 5 6 4 " 0 . 0 5 4 8 " '( - 4 . 0 6 ) ( 4 . 3 2 )

0 .0176 - - 0 .0133 . '( 4 .08 ) ( 3 .34 )

- . 028s . .( -4 .42 t

0 .4386 ' . 0 . 4369"(3 .63 ) ( 4 .04 )

0.2545" 0 .2584"(2 . s0 ) ( 2 .83 )

0 .1448 0 .1553(0 .5e ) ( 0 .7 r_ )

- . 0004 - . 0038( -o .42 ) ( -o .64 )

- . 0002 - . 0001( -0 .36 ) ( - 0 .10 )

- . 0064 - - . 0068 "( -2 .44 ) ( - 2 .8e )

R: .83s6 .87ssD.H . 0 .0085 -2 .4568 - 's . s .E . . 00380 . 00320Q(24) 20 .3 r l - 7 . 00

.8631 .8902 .8634 .8903-o .6L96 -0 .9549 -0 .6188 1 . 1001

.00352 .00280 .00351 . o027 It - 8 .85 L3 .44 18 .71 L3 .34

8 8 : L - 9 4 : 4Forecasts . s . E . . 0 0 1 7 0 . 0 0 1 - s 6 . 0 0 1 3 3 . 0 0 L 2 2 . 0 0 1 3 2 . 0 0 1 2 1' ( " ) - -s ign i f icant at the 5t (1- t ) 1evel . t - ra t ios are in parentheses.

Var iable Def in i t ions:EC = error correction term for bank loan outstandings.CR = index of change in bank willingness to nake consumer l"oans,Lb = real stock of U.S. bank consumer loans, securi t-i-zaxion adjusted.CON = proxy for peruanent income.DUM802 = credit control dumrny : I in 1980:Q2.CON = MPS rnodel's proxy for permanent income.SPREAD = finance conpany auto loan rate minus bank auto loan rate, new cars.t = real . a f ter - tax auuo loan rate.U E civilian unenplo)rment rate.

ob0

:

c) c

.) \o

a

.. o

I

F- \t

.-r -C

..

r?

>N

N

,-.1

oO

t

'-{o\

\+

o .o

o t

o q-l o1

'oI

o-

5 "

,,ix

d

q.'l'_

{ oal '''

! q

CC

r

-.1

'C

l -

tru

..f, o

.r N

C

.-r I

!dN

oi\

rl

,-llP

H!(

D0

1

F.{

O

'rr O

tr'E

lo

ttt d

F

. tr

O...1

I d

o

r>g

'iog

I9N

H

'l O

.

oo

o

''l a

)o

@

p

o

lfo|1 F

- r-t

r-l <

h O

$

O

d\t

\a <

. -O

r-.r

O

bO tr

+O

r-l

Ii I

A

or-.1 F

\ o

! d

q-l u

c

. o

J

.,.r lr

o

>\.,r

.t .o

(u

P

!-l

q

O ''l

Otr

f O

O

'rl

d) ll

.rr .ri

.uq

lll

lrlJ

d

3x

o

o;

o

Ea

\ q

o

!o:

.-r C

E

Y,.rP

co

r^

f.. o

1

9q

9

lt--

N

F.t

(J!

qr3

3

o\N

O

"

H

o

.c

o

. o

o

-r lr

o€

.N

u

o

lll |{J

tr

.- c

, ^

in p

!-+,

o

EE

sH

EE

d

o+

,E

lro

.i( ).,1

o\o

lrrJ3

.oo

..1

\oo

o

..ro9

'H!

rfl r.r +

d

.Clt

JO

r ..{

C\l

O

O, !

'\'O

(!|-r

.O

! !

A

9?

. c

!. to

xx

.pq

)

lll {,

rrl.,r

'd a

Io

J

o

X

H

rrq

.rod

Ho

oF

d "'i -_

(, rt'l

co

o

r E

H

>N

\o

$

!z"l

"' o

'd.-.{

l'\ $

o

O'd

E

g

o

o\

ON

rJ

(J

X'H

lJ

r-- .o

h

ox

uJ

' N

'

l -c

F

q

a.

OO

'rr.:lrE.-r E

'E'

I o

pd

F.t!c

)

O r-r trr^

: q

,

r-{ \o

!t.|

:: q

)o

$

co

o

o

o

x

l

po

\ .-r o

O

O

E.-.1

X

o

O

rrl r-r c

\ 5

.o

;.-r E

r- .o

!

o

(!.:i.

c{

O C

tr;^E

l.r o

J -

oP r-r

r. o€

J-r

Jo

d;'i

ot il

o

i:dp

rd p

-ln

o

|.r\ rn

o

Ea

n

oo

.n

c{

rrl 3

"'

o.,1

\o

r-r N

O.n

o

E

+r

(n

o\ d

\ .'l lr6

>

\u

or-\

'O

ol

6

+.1

. r-r

. C

l o

.-P!

Ol

o C

ut

(! d

.dl (D

o

o .u

lJl C

..l .rD

o,rl

h0 +l

_o tr

flc

6

d

a.rll

.rl .r.{

tr O

-tr

E

q-rl,-.t

(J J

:.1g

?

ol'-r

o€

H!

! O

ltl lr

'i(.)

15

o

.^

tr o)

o

ol

o

o

bo

o!

Frl ta

d

!.-r d

tr .

-Ol d

J

o

.'ro

- ld

d

l (d lll d

rh

(, ,

. .

.,rl _o d

^""c

F

(^

rrl J

o

i{n

t tr\ trv

ld€

€v

) >

l '.{ "r -* '-r

_.f,o

r :

. F

.{

.'IO

V

i^

i r^

cf)o

u

1^

\o

\oO

!.1

O

!.1 oO

O\

(\

. J

cd

Ou

l O

O

\ oO

'| .N

'N

lv

Ov

O

v

\olI

ao

^

N^

r-rl N

F

- o

.-.1O

l -i

.J .-l

+E

l o

o

=l

o-

c,J

-f, ^

\a r'1

Fr

r-.1 co

O

O

ON

.o

Ic

rv

lv

6I^

N

O

. r-l

. I

ov

lv

J^

J

O

ON

(.

.-l .

O

r-l.N

I

ov

I

v

:^

" I

i ,

O\ \O

N

^

\O^

O

\^

O^

, !n

i^

Ot--

or cQ

(Y

) tr)

o1 o\

r-r o\

!n o

dl :.,j

(l1O

\ .

Nd

l O

\ \O

O-J

lO

N

c\

. ;

r-r \t

O

N

' ..)

' -f,

. 6

! c.J

;iN.l

.(\

.N

.rn

.c

{

(r

l |

'i I

lv

ov

c

)v

c

)v

o

v

t'v

I v

ii\o^

c

r^

rJJ

^

c{@

N

+O

\O

c'\

O\

fO

O1

ritrr

|Jt\o

o\ \o

r-.{ ltl

\o f-i

rft .

or .

N

. .rl

. $

(.

c.l

ON

,<\

'.i .N

|

|C

)v

O

v

Ov

lv

lv

I^N\O.l

lv

Oi^

r

-{

f-^

\O

^

(E

N

ON

\o \o

o

o\

\o4

.a

+

F-+

a

- co

d\ !O

c.l

N

(n ir

oO

\O

@

. \O

.

O

rr c.,l

c) r-l

r-l O

r-l

'O

.4

.N

' r.l

| |

ov

o

v

c'v

c

'v

lv

lv

+ov

l^o rr_r

lv

Lt)lI o

\ \t $

^rrl

co (n

\t N

{Jl F

r J

lrl\Jl

O

r-r O

ot

| .o

=l

tv

c'v

l^

rr co

s

NO

N

U.

.O

6

\+.o

.c

{O

V

OV

+l

i^I

l.- \i

N

cf) F

.l o\

r-rl .n

F.1

oO \O

O

\t

ol -r

F.r

o

ot

| |

|tl

c,v

lv

tv

i^FC

{It-

O\

Or

..(\O

v

IE

IOl

c.l O

|.-

F-

=l

rrt N

(a-'

l.-{\o.

4l

oJ

O

c{

&tll

hl

lv

rv

..1| \a

r- dt

(aF

rl <

E cn

4 \O

Ol

r-r F

- .

ot

| .l

=l

rv

rv

NII

ea

$

oo

rafrl

r.- N

o\

ra)O

l F

.] \O

.

!l O

e{

O

r-lo

t tl

=l

lv

rv

r'rl I

| \o

co

F

- o\

(a^

Frl

ql e-{

\o N

\o

NO

l N

O

q

\@5l

o c.I

i-l (\

oo

l |

| .N

5l

tv

tv

tv

'I's

lo

l N

$

.''NE

l r-1 co

\+

..)l\o

N.

4l

o\t

N\+

&lll

trrl lv

lv

oo, ,^

,-r qp

r-r >

6z

o0)b0od{)-o'd

.lr

oii ut

<';

'dH

ri<

.-l!1!,)

i I

i^(r"\ ^

ul^

$

^

NJ

9

0J

C)

N

N

\tN

\t C

!d

.n

\O

cO

O

l'\

6

.\i

N

. d

) .

Nc{

F

(r','+

.N

' r-r

| .

lc

)v

o

v

ov

I

v

-i()AE.r(

tulr\<

Ei-|.a

r

\\L;X

$rr-l

.-l ..1

1n

v()

-r

l]

".*

xo

&,

*ia

-_

! 9

Z

thiS

q)d

.-,'

{ll !

!l til

El

u,..1

1

O

d!l

F;

qo

()>

t o

El

Table 6: Selected

Variables

) i=o. zY'r-r

otL

E1=o , , (Y I *CR) . , '

Er=', , (yI*CONTROL).-t

32

Sta t is t i cs From

FRBModel

0 .7013 . .( 10 .0s )

L .2180 - -(8 .40 )

0 .2556"(2 .83 )

0 . 0291 '( 2 . s8 )

0 .0s94 . .( 4 .7e )

- . 0500( -1 . e0 )

CoN Regress ions

Model 7

0 . 1 2 3 5 "( r3 .2L )

t .0424. '( 7 . 44 )

0 .2588 " "(3 .14 )

0 .032s "(3 .16 )

0 .o62L "(5 .7 r )

- . 0570 '( - 2 .20 '

0 . 037 3' . -( 2 .88 )

(1969 : Q1-94 :Q4)

Model 8

0 . 7249 "(L2 .7L )

1 .0750 - '( 7 .38 )

o .2s97 "(3 .10 )

0 .0294"(2 . 11 t

0 . 0611"( s .46 )

- . 0541 '( - 2 .08 )

0 .0393 . '( 2 .77 )

0 .0037(0 . L7 )

R, .99957rho .9267 4D .W. L .75s . s . E . . I I 37 6- ( - - ) - -s ign i f icant at the 5g (1*) level .

Var iable Def in i t ions:CON = MPS model's proxy for permanentyt = real labor income.y" = real income from transfers.yp = af ter - tax proper ty income (e.g.ws = the six-quarter noving averaged' = the six-quarter movlnB averageOIL = 1974 oLL shock duxony.

, d iv idends, in teres! , and rent) .of real stock narket wealth.of real non-stock narket r,/ealEh.

. 9 9 9 5 5

. 9 0 1 6 3L . 9 2.L0457

t- rat ios

income.

. 9 9 9 5 5

.90699L . 9 6. to224

are in parenuheses.

3 3

List of Variable Definitions (Not intended for publication)

Ei = expected loan revenue from household i,q t = expected loan qual i ty of household i .d1 = expected default rate of household i.y' = underlying incone potential of household i.6 , i = real , r isk less cost of funds to banks.P = positive indicator of the rnacroeconomic outlook.f = notional loan dernand.L' = average per capita dernand for real consumer loans from all lenders.Lb = real s tock of U.S. bank consumet 1oans, secur i t izat ion adjusted.so = bank share of total consumer loans.a = population share tha! qualifies for a loan and botrows.N = populat ion.C = real personal consumption expenditures.r = real, tax-adjusted bank auto loan rate.tr" = expectations of inflation.R = nominal bank auto loan rate (source: Federal Reserve surveys)t = rnarginal federal income tax rate for a proto-t)rpical fanily of four.o = cutrent share of consumer loan interest that is tax deductible,O = 4-year ahead share consumer loan intetest that is tax deductible,DUR = real consumer durable spending growth ($ 1987).cR = index of change in bank willingness to make consumer loans.A = f i rs t d i f ference operator .Al" = percent change in real bank consumer loan outstandlngs.AIeb = percent change in real bank consuuer installment loan extensions.ARFF = first difference of the real federal funds rate.GM22 = 2 quarter growth rate of real M2.A?APERBILL = A of 4-6 rnon. cornmercial paper rate ninus 6 nonth T-bill rate.AYCURVE = A of che 10-yr. Treasury note-3 month T-bill spread.CLI = 1 qtr. percent change in the index of leading economic indicators.GLI2 = 2 qtr. percent change in the index of leading economlc indicators.REGQ = neasure of bindingness of regulatlon Q ceilings.CoNTROL = credit control durnroy - 1 in 1980:2 and -l in 1,980:3.MMDA = dunny equal to 1 in 1982:Q4 when MMDAs were introduced.EC = appropriate error correction terms fot various models.DUM802 = credit control dunny : 1 in 1980:Q2.CON = MPS rnodel's proxy for permanent income.SPREAD = finance company auto loan rate minus bank auto loan rate, new cars.r - rea l , a f ter - tax auto loan rate.U = ciwllian unemploynent rate.L/k = depreciation adjusted durable spending divided by the durable stock.r ' ( r ' ) = user cost of capi ta l for auto (nonauto) durables, deprec iat ion adj .yu = real d isposable income.[Pg4/P@"*MPG] = real prlce of gasoline per car rniLe driven.Psas = nominal price index for gasoline.

= price deflator for CoN, the MPS model's measure of permanent incorne.MPC = gas rnileage adjustnent for calculating the real price of gasollne

Per caf ml- .Le o l lven.yt = real labor ineome.

= real income from transfers.yn = af ter - tax proper ty incorne (e.g. , d iv idends, in terest , and rent) .w" = the six-quarter noving average of real stock market wealth.!r' = the six-quarter moving average of real 494-stock narket wealth.OIL = L974 oL]- shock durnmy.

RESEARCH PAPERS OF THE RESEARCH DEPARTMENTFEDERAL RESERVE BANK OF DALLAS

Available, at no charge, ftom the Re.search DepartrnentFederal Reserve Bank of Dallas, P. O. Box 655906

Dallas. Texas 7 5265-5906

Please check the titles of the Re$earch Papers you would like to receive:

92Ol Are Deep Recessions Followed by Strong Recoveries? (Mark A. Wynne and Nathaa S. Balke)92OZ The Case of the "Missing M2" (Jobn V. Duca)9203 Immigrant Links to the Home Country: Implications foi Trade, Welfare and Factor Rewards

(David M. could)9204 Does Aggregate Output Have a Unit Root? (Mark A. Wynne)9205 hflation and lts Variabilify: A Note (Kenneth M. Emery)9206 Budget Co$trained Frontier Measures of Fiscal Equality and Efficiency in Schooling (Shawna

Grosskopf, Kathy Hayes, Lori L. Taylor, William \ryeber)9207 The Effects of Credit Availability, Nonbank Cornpetition, and Tax Reform on Bad< Consumer

Ifiding (John V. Duca and Bonnie Garren)9208 On the Future Erosion of the North American Fre€ Trade Agreement (Wi iam C. Gruben)9ZO9 Threshold Cointegation (Nathan S. Balke and Thomas B. Fomby)92lO Cointegration and Tests of a Classical Model of Inflation in Argentina, Bolivia, Brazil, Mexico, and

Peru (F.aul Anibal Feliz and John H. Welch)mll Norninal Feedback Rules for Monetary Policy: Some Comments (Evan F. Koenig)mQ The Analysis of Fiscal Policy in Neoclassical Models' (Mark Wynne)9213 Measuring the Value of School Qualiry (Lori Taylor)y2l4 Forecasting Turning Points: Is a Two-State Characterization of the Business Cycle Appropriate?

(Kemeth M. Ernery & Evan F. Koenig)9215 Energy Security: A Comparison of Protectionist Policies (Mine K. Yiic€l and Carol Dahl)9216 An Amlysis of rhe Impact of Two Fiscal Policies on the Behavior of a Dynamic Asset Market

(Gregory W. Huftoaa)9301 Human Capital Extemdities, Trade, and Economic Growth (David Gould and Roy J. Ruffin)93OZ The New Face of Latin America: Financial Flows, Markets, ard Institutions in the 1990s (John

Welch)9303 A General Tlvo Sector Model of Endogenous Growth with Hunan and Physical Capital (Eric Bond,

Ping Wang, and Chong K. Yip)9304 The Political Economy of School Reform (S. Grosskopf, K. Hayes, L. Taylor, and W. Weber)9305 Money, Output, and Income Velocity (Theodore Palivos and Ping Wang)9306 Constructing an Alternative Measure of Changes itr Reserve Requirement Ratios (Joseph H. Haslag

and Scott E. Hein)9307 Money Demand and Relative Prices During Episodes of Hyperinflation (Ellis W. Tallman and Ping

w*g)930E On Quantity Theory Restrictions and the Signalling Value of the Money Multiplier (Joseph Haslag)9309 The Algebra of Price Stability (Nathan S. Balke and Kenneth M. Emery)9310 Does It Matter How Monetary Policy is Implemented? (Joseph H. Haslag and Stott Hein)931 I Real Effects of Money and Welfare Costs of Inflation in an Endogenously Growing Economy with

Transactions Costs (Ping Wang and Chong K. Yip)9312 Borrowing Constraints, Household Debt, a:rd Racial Discrimination in Loan Market$ (John V. Duca

and Stuart Rosenthal)9313 Default Risk, Dollarization, and Cunency Substitution in Mexico (Williarn Gruben and John Welch)9314 Tecbnological Unemployment (W. Michael Cox)9315 Output, Inflation, and Stabilization in a Small Open Economy: Evidence from Mexico (Joh H.

Rogers and Ping Wang)9316 Price Stabilization, Output Stabilization and Coordinated Monetary Policy Actions (Jos€ph H.

Haslag)9317 An Altemative Neo{lassical Growth Model with Closed-Form Decision Rules (Gregory W.

Huffnan)9318 Why the Composite Index of I*ading Indicators Doesn't l€ad (Evatr F. Koedg and Kenneth M.

E "q')

9319 Allocative Inefficiency and I-ocal Government: Evidence Rejecting the Tiebout Hpothesis (Lori L.Taylor)

9320 The Output Effects of Govemment Consumption: A Note (Mark A. Wynne)9321 Should Bond Funds be Included in M2? (John V. Duca)9322 Recessions and Re.ove es in ReaI Business Cycle Models: Do Real Business Cycle Models

Generate Cyclical Behavior? (Mark A. WyDne)9323* Retaliation, Liberalization, ald Trade Wan: The Political Economy of Nonstrategic Trade Policy

(David M. Gould and Graeme L. Woodbridge)9324 A General Two-Sector Model of Endogenous Growth with Human ald Physical Capital: Balanced

Grorvth atrd Transitional Dynamics (Eric W. Bond, Ping Wang,a.nd Chong K. Yip)9325 Gmwth and Equity with Endogenous Human Capital: Taiwan's Economic Miracle Revisited (Maw-

Lin L€e, Ben-Chieh Liu, and Ping Wang)9326 Clearinghouse Banks aad Banknote Over-issue (Scott Freeman)9327 Coal, Natural Gas and Oil Ma.rkets after World War II: Wtrat's Old, What's New? (Mine K. Yiicel

ard Shengyi Guo)9328 On the Optimality of Interest-Bearing Reserves in Economies of Overlapping Cenerations (Scott

Freeman and Joseph Haslag)9329+ Retaliation, Liberalization, and Trade Wars: The Political Economy of Nonstrategic Trade Policy

(David M. Gould and Graeme L, Woodbridge) (Reprint of 9323 h error)9330 On the Existence of Nonoptimal Equilibria in Dynanic Stochastic Econonies (Jeremy Greenwood

and Gregory W. Huffman)9331 The Credibility and Performance of Unilateral Target Zones: A Comparison of the Mexican and

Chilean Cases (Raul A. Feliz and John H. Welch)9332 Endogenous Growth and Intemational Trade (Roy J. Ruffin)9333 Wealth Effects, Heterogeneity and Dynamic Fiscal Policy (Zsolt Becsi)9334 The Ilefficiency of Seigniorage ftom Required Reserves (Scott Freeman)9335 Problems of Testitrg Fiscal Solvency in High Inflation Economies; Evidence from Argentina, Brazil,

and Mexico (John H. Welch)9336 Income Taxes as Reciprocal Tariffs (W. Michael Cox, David M. Gould, and Roy J. Ruffin)9337 Assessing the Economic Cost of Unilateral Oil Consewation (Stephen P.A. Brown and Hillard G.

Huntington)9338 Exchange Rate Uncenainty and Economic Growrl in Latin Anerica (Darryl Mcleod and Jobn H.

Welch)9339 Searching for a Stable M2-Demand Equation (Evan F. KoenD9340 A Suwey of Measurement Biases in Price Indexes (Mark A. Wynne and Fiona Sigalla)9341 Are Net Discount Rates Stationary?: Some Further Evidence (Joseph H. Haslag, Michael

NieswiadomJ, and D. J. Slottje)9342 On the Fluctuations Induced by Majority Voting (Gregory W. Huffman)9401 Adding Bond Funds to M2 in the P-Star Model of Inflation (Zsolt Becsi atrd John Duca)9402 Capacity Utilization and the Evolution of Manufacturing Output: A Closer Look at the "Bounce-

Back Effect' (Evan F. Koenig)9403 The Disappearing January Blip and Othei State Employment Mysteries (Frant Berger and Keith R.

Phillips)9404 Energy Policy: Does it Achieve its Lntended Goals? (Mine Yiicel ard Shengyi Guo)9405 Protecting Social Interest in Free Invention (Stephen P.A. Brown and William C. Gruben)9406 The Dynamics of Recoveries (Nathan S. Balke and Mark A. Wynne)9407 Fiscal Poliry in More General Equilibriium (Jim Dolman and Mark Wynne)9408 On the Political Economy of School Deregulation (Shawna Grosskopf, Kathy Hayes, Inri Taylor,

a:rd William Weber)9409 The Role of Intellectual Property Rights in Economic Growth (David M. Gould and Willia:l C.

Gruben)9410 U.S. Banks, Competition, and the Mexican Banking System: How Much Will NAFTA Matter?

(William C. Gruben, Jobn H. Welch and Jeffery W. Gunther)94ll Monetary Base Rules: The Cunency Caveat (R. W. Hafer, Joseph H. Haslag, andscott E. Hein)94L2 The Information Content of the Paper-Bill Spread (Kenneth M. Emery)9413 The Role of Tax Policy in the Boom/Bust Cycle of the Texas Construction Sector (D'Ann Petersen,

Keith Phiuips and Mine Yiicel)94L4 The P* Model of Inflation, Revisited (Evan F. Koenie)9415 The Effects of Monetary Policy in a Model with Reserve Requirements (Joseph H. Haslag)

9501 An Equilibrium Analysis of Central Bank Independence and Inflation (Gregory W. Huffman)9502 Inflation and lntermediation in a Model with Endogenous Growrh (Joseph H. Haslag)9503 Country-Bashing Tariffs: Do Bilateral Trade Deficits Matter? (W. Michael Cox and Roy J. Ruffrn)9504 Building a Regional Forecasting Model Utilizing Inng-Term Relationships and Short-Term

Indicators (Keith R. Phillips and Chih-Ping Chang)9505 Building Trade Barriers and Knocki.ng Them Down: The Political Economy of Unilateral Trade

Liberalizarions (David M. Gould and Graeme L. Woodbridge)9506 On Competition and School Efficiency (Shawna Grosskopf, Kathy Hayes, Inri L. Taylor and

William L. Weber)9507 Alternative Methods of Corporate Confol i.n Commercial Banks (Stephen Prowse)9508 The Role of Infatemporal Adjustment Costs in a Multi-Sector Economy (Gregory W. Huffiran

and Mark A. Wyme)9509 Are Deep Rec€ssions Followed By Strong Recoveries? Results for the G-7 Countries (Nathan

S. Balke and Mark A. Wynne)9510 Oil Prices and Inflation (Stephen P.A. Brown, David B. Oppedahl and Mine K. Yiicel)9511 A Comparison of Altemative Monetary Environments (Joseph H. Haslag))9517 Regulatory Changes and Housing Coefficients (John V. Duca)9513 The lnterest Sensitivity of GDP and Accurate Reg Q Measures (Job:r V. Duca)9514 Credit Availability, Bank Consumer Irnding, and Consumer Durables (Joh.n V. Duca and

Bonnie Garrett)

Name: Organization:

Address: City, State and Zlp Code:

Please addmeto your mailing list to receive future Research Papers: _ Yes _No